robert1003 / adl-hw2

ADL hw2

Homework description

• Link to the homework slide.
• Link to the homework video.
• Link to the data.
• Link to the evaluation script.

Execution details

Train

1. create directory final/model
2. run python3.7 final/train.py [train_data_path]
3. trained model will be stored in final/model

Test

1. download pretrained model file with bash download.sh
2. run bash run.sh [test_data_path] [predict_path]

Plot Answer Length Distribution (Q5)

1. run python3.7 final/draw_p5.py [train_data_path] [output_pic_path]

Plot Answerable Threshold (Q6)

1. pretrained model file should be placed in final/model
2. run python3.7 final/draw_p6.py [dev_data_path] [output_pic_path] [python3 path (installed with ckiptagger package)] [scorer_path] [temp_file_1_path] [temp_file_2_path] [ckipdata_path]
   • temp_file_1_path and temp_file_2_path represents two temporary files